Fuzzy Timed Agent Based Petri Nets for Modeling Cooperative Multi-Robot Systems

Xingli Huang, Hua Xu, Peifa JiaInt. J. Commun. Netw. Syst. Sci 2009 · OpenReview

Abstract

A cooperative multi-robot system (CMRS) modeling method called fuzzy timed agent based Petri nets (FTAPN) is proposed in this paper, which has been extended from fuzzy timed object-oriented Petri net (FTOPN). The proposed FTAPN can be used to model and illustrate both the structural and dynamic aspects of CMRS, which is a typical multi-agent system (MAS). At the same time, supervised learning is supported in FTAPN. As a special type of high-level object, agent is introduced into FTAPN, which is used as a common modeling object in its model. The proposed FTAPN can not only be used to model CMRS and represent system aging effect, but also be refined into the object-oriented implementation easily. At the same time, it can also be regarded as a conceptual and practical artificial intelligence (AI) tool for multi-agent systems (MAS) into the mainstream practice of the software development.

Introduction

Cooperative Multi-Robot Systems (CMRS) are essentially complex Multi-Agent Systems (MAS) where robots must communicate, synchronize, and make autonomous decisions under strict time constraints. Modeling these systems is notoriously difficult. Standard Petri Nets (PN) are excellent for modeling concurrency but lack the ability to describe time. Timed Petri Nets (TPN) add temporal dimensions but often become mathematically "explosive" and unmanageable as system complexity grows. While Object-Oriented Petri Nets (OPN) help organize complex systems into manageable objects, they generally fail to model learning or "aging effects"—where system performance degrades or information relevance changes over time.

Researchers from Tsinghua University introduced Fuzzy Timed Agent Based Petri Nets (FTAPN) to address these gaps. FTAPN serves as a "bridge" that allows designers to move from a conceptual multi-agent model to a practical, object-oriented software implementation without losing the ability to analyze temporal constraints and fuzzy reasoning.

The FTAPN Framework

FTAPN extends previous work on Fuzzy Timed Object-Oriented Petri Nets (FTOPN) by introducing the "Agent" as a first-class modeling object. This allows for a more direct translation of the model into Object-Oriented Concurrent Programming (OOCP).

Agent Autonomy via ACTALK

To ensure true autonomy rather than just reactive behavior, the framework integrates the ACTALK model for "active objects." This provides every robot (agent) with its own address, a mailbox for asynchronous messages, and an internal activity thread. This structure ensures that agents can manage their own state and respond to environmental stimuli independently.

The ACTALK model for active objects, providing the foundation for agent autonomy in FTAPN.

Fuzzy Firing Logic and Temporal Decay

Unlike traditional nets where a transition either fires or doesn't (binary 0 or 1), FTAPN uses fuzzy t-norms and s-norms to calculate a continuous "level of firing." This allows the model to handle uncertainty and partial truths, which are common in real-world robotic environments.

Furthermore, the model incorporates an exponential decay function for tokens to represent the "aging effect." This accounts for the diminishing relevance of old sensor data or the physical wear of robotic components over time. The temporal constraints are represented through fuzzy intervals, as shown in the various transition types below.

Fuzzy timed transitions in FTAPN, illustrating different temporal constraints and intervals.

Supervised Learning

A key innovation of FTAPN is its support for supervised learning. The net includes tunable parameters such as weights ($w_i$), thresholds ($r_i$), and temporal decay factors ($\alpha_i$). Using a gradient-descent-style update formula, the model can be "trained" using experimental input/output data to minimize error, allowing the system to adapt its behavior based on past performance.

Validation and Results

The authors validated the FTAPN method using an industrial case study: a three-robot wafer etching platform used in semiconductor manufacturing. In this system, one robot handles atmospheric-to-vacuum transfer, while two others manage internal vacuum tasks.

By feeding the learning algorithm discrete time-series data representing the "markings" (token values) of input and output places, the researchers were able to estimate the net's internal parameters accurately. Comparative analysis demonstrated that FTAPN covers dynamic learning, autonomous decision-making, and transparent knowledge representation more effectively than standard MAS or FTOPN frameworks alone.

Conclusion

FTAPN offers a robust "Glass Box" modeling environment where the internal logic of a problem is mapped directly onto the net topology, making the system's reasoning transparent to the designer. While the framework successfully models CMRS and supports refinement into practical AI tools, challenges remain. Specifically, state-space analysis is difficult due to the complexity of fuzzy and temporal variables. Future work will focus on developing Extended State Graphs and better incorporating certainty factors for firing in highly unpredictable environments.

Summary updated

Abstract

A cooperative multi-robot system (CMRS) modeling method called fuzzy timed agent based Petri nets (FTAPN) is proposed in this paper, which has been extended from fuzzy timed object-oriented Petri net (FTOPN). The proposed FTAPN can be used to model and illustrate both the structural and dynamic aspects of CMRS, which is a typical multi-agent system (MAS). At the same time, supervised learning is supported in FTAPN. As a special type of high-level object, agent is introduced into FTAPN, which is used as a common modeling object in its model. The proposed FTAPN can not only be used to model CMRS and represent system aging effect, but also be refined into the object-oriented implementation easily. At the same time, it can also be regarded as a conceptual and practical artificial intelligence (AI) tool for multi-agent systems (MAS) into the mainstream practice of the software development.
BibTeX
@article{huang2009fuzzy,
  author = {Xingli Huang and Hua Xu and Peifa Jia},
  title = {{Fuzzy Timed Agent Based Petri Nets for Modeling Cooperative Multi-Robot Systems}},
  journal = {Int. J. Commun. Netw. Syst. Sci.},
  year = {2009},
  url = {https://openreview.net/forum?id=CcGCg2e87Mi},
}
Figure